Unread oracle bone charactersEvidence
Evidence · Unread oracle bone characters
A 2020 deep-learning system ranked the correct join in its top 10 for 98.39 percent of held-out fragment pairs, and…
A 2020 deep-learning system ranked the correct join in its top 10 for 98.39 percent of held-out fragment pairs, and platform tools have since found about 500 duplicate rubbings.
What it means
Key questionHow many of the 150,000 fragments can still be rejoined, and how much context does rejoining recover for rare graphs?
No answer ahead yet
Openwas No evidence yet
Answer this evidence bears onAutomated rejoining recovers context at scale
Supports1 for, none against
Still possibleunchanged
Next steps 1
1
Watch for the results of the 2025 Anyang AI competition (due late October 2025) and record the number of algorithm-proposed joins confirmed by palaeographers.
~3 h
—not judged
Ranked by how likely each is to move the key question, for the time it takes
Sources 2
SRC
SRC
Cite it
Mysteries Department. (2026). A 2020 deep-learning system ranked the correct join in its top 10 for 98.39 percent of held-out fragment pairs, and… [Evidence EVD0DEDTM293ZBW]. https://mysteriesdept.com/r/EVD0DEDTM293ZBW
@misc{mysteriesdept_evd0dedtm293zbw,
title = {A 2020 deep-learning system ranked the correct join in its top 10 for 98.39 percent of held-out fragment pairs, and…},
author = {{Mysteries Department}},
year = {2026},
note = {Evidence, catalog key EVD0DEDTM293ZBW},
url = {https://mysteriesdept.com/r/EVD0DEDTM293ZBW}
}